Btrfs: Seed device support

Seed device is a special btrfs with SEEDING super flag
set and can only be mounted in read-only mode. Seed
devices allow people to create new btrfs on top of it.

The new FS contains the same contents as the seed device,
but it can be mounted in read-write mode.

This patch does the following:

1) split code in btrfs_alloc_chunk into two parts. The first part does makes
the newly allocated chunk usable, but does not do any operation that modifies
the chunk tree. The second part does the the chunk tree modifications. This
division is for the bootstrap step of adding storage to the seed device.

2) Update device management code to handle seed device.
The basic idea is: For an FS grown from seed devices, its
seed devices are put into a list. Seed devices are
opened on demand at mounting time. If any seed device is
missing or has been changed, btrfs kernel module will
refuse to mount the FS.

3) make btrfs_find_block_group not return NULL when all
block groups are read-only.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>

8 files changed, 932 insertions, 315 deletions

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 8bb452456d9..dd1c03aea2d 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -185,6 +185,10 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	btrfs_set_header_owner(cow, new_root_objectid);
 	btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
 
+	write_extent_buffer(cow, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(cow),
+			    BTRFS_FSID_SIZE);
+
 	WARN_ON(btrfs_header_generation(buf) > trans->transid);
 	ret = btrfs_inc_ref(trans, new_root, buf, cow, NULL);
 	kfree(new_root);
@@ -274,6 +278,10 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	btrfs_set_header_owner(cow, root->root_key.objectid);
 	btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN);
 
+	write_extent_buffer(cow, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(cow),
+			    BTRFS_FSID_SIZE);
+
 	WARN_ON(btrfs_header_generation(buf) > trans->transid);
 	if (btrfs_header_generation(buf) != trans->transid) {
 		u32 nr_extents;
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index c4c6c127323..5ff74282a62 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -177,6 +177,9 @@ struct btrfs_dev_item {
 	/* type and info about this device */
 	__le64 type;
 
+	/* expected generation for this device */
+	__le64 generation;
+
 	/* grouping information for allocation decisions */
 	__le32 dev_group;
 
@@ -188,6 +191,9 @@ struct btrfs_dev_item {
 
 	/* btrfs generated uuid for this device */
 	u8 uuid[BTRFS_UUID_SIZE];
+
+	/* uuid of FS who owns this device */
+	u8 fsid[BTRFS_UUID_SIZE];
 } __attribute__ ((__packed__));
 
 struct btrfs_stripe {
@@ -263,6 +269,7 @@ struct btrfs_header {
 					sizeof(struct btrfs_item) - \
 					sizeof(struct btrfs_file_extent_item))
 
+#define BTRFS_SUPER_FLAG_SEEDING (1ULL << 32)
 
 /*
  * this is a very generous portion of the super block, giving us
@@ -278,7 +285,7 @@ struct btrfs_header {
 struct btrfs_super_block {
 	u8 csum[BTRFS_CSUM_SIZE];
 	/* the first 4 fields must match struct btrfs_header */
-	u8 fsid[16];    /* FS specific uuid */
+	u8 fsid[BTRFS_FSID_SIZE];    /* FS specific uuid */
 	__le64 bytenr; /* this block number */
 	__le64 flags;
 
@@ -941,6 +948,7 @@ BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64);
 BTRFS_SETGET_FUNCS(device_group, struct btrfs_dev_item, dev_group, 32);
 BTRFS_SETGET_FUNCS(device_seek_speed, struct btrfs_dev_item, seek_speed, 8);
 BTRFS_SETGET_FUNCS(device_bandwidth, struct btrfs_dev_item, bandwidth, 8);
+BTRFS_SETGET_FUNCS(device_generation, struct btrfs_dev_item, generation, 64);
 
 BTRFS_SETGET_STACK_FUNCS(stack_device_type, struct btrfs_dev_item, type, 64);
 BTRFS_SETGET_STACK_FUNCS(stack_device_total_bytes, struct btrfs_dev_item,
@@ -960,12 +968,19 @@ BTRFS_SETGET_STACK_FUNCS(stack_device_seek_speed, struct btrfs_dev_item,
 			 seek_speed, 8);
 BTRFS_SETGET_STACK_FUNCS(stack_device_bandwidth, struct btrfs_dev_item,
 			 bandwidth, 8);
+BTRFS_SETGET_STACK_FUNCS(stack_device_generation, struct btrfs_dev_item,
+			 generation, 64);
 
 static inline char *btrfs_device_uuid(struct btrfs_dev_item *d)
 {
 	return (char *)d + offsetof(struct btrfs_dev_item, uuid);
 }
 
+static inline char *btrfs_device_fsid(struct btrfs_dev_item *d)
+{
+	return (char *)d + offsetof(struct btrfs_dev_item, fsid);
+}
+
 BTRFS_SETGET_FUNCS(chunk_length, struct btrfs_chunk, length, 64);
 BTRFS_SETGET_FUNCS(chunk_owner, struct btrfs_chunk, owner, 64);
 BTRFS_SETGET_FUNCS(chunk_stripe_len, struct btrfs_chunk, stripe_len, 64);
@@ -1661,6 +1676,7 @@ int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
 			       struct extent_buffer *buf, u64 orig_start);
 int btrfs_add_dead_reloc_root(struct btrfs_root *root);
 int btrfs_cleanup_reloc_trees(struct btrfs_root *root);
+u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags);
 /* ctree.c */
 int btrfs_previous_item(struct btrfs_root *root,
 			struct btrfs_path *path, u64 min_objectid,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c599f0ee997..82833e5d84b 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -345,6 +345,25 @@ out:
 	return 0;
 }
 
+static int check_tree_block_fsid(struct btrfs_root *root,
+				 struct extent_buffer *eb)
+{
+	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
+	u8 fsid[BTRFS_UUID_SIZE];
+	int ret = 1;
+
+	read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
+			   BTRFS_FSID_SIZE);
+	while (fs_devices) {
+		if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
+			ret = 0;
+			break;
+		}
+		fs_devices = fs_devices->seed;
+	}
+	return ret;
+}
+
 int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 			       struct extent_state *state)
 {
@@ -382,9 +401,7 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 		ret = -EIO;
 		goto err;
 	}
-	if (memcmp_extent_buffer(eb, root->fs_info->fsid,
-				 (unsigned long)btrfs_header_fsid(eb),
-				 BTRFS_FSID_SIZE)) {
+	if (check_tree_block_fsid(root, eb)) {
 		printk("bad fsid on block %Lu\n", eb->start);
 		ret = -EIO;
 		goto err;
@@ -1558,9 +1575,11 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
-	err = btrfs_parse_options(tree_root, options);
-	if (err)
+	ret = btrfs_parse_options(tree_root, options);
+	if (ret) {
+		err = ret;
 		goto fail_sb_buffer;
+	}
 
 	/*
 	 * we need to start all the end_io workers up front because the
@@ -1610,18 +1629,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_start_workers(&fs_info->endio_write_workers,
 			    fs_info->thread_pool_size);
 
-	err = -EINVAL;
-	if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) {
-		printk("Btrfs: wanted %llu devices, but found %llu\n",
-		       (unsigned long long)btrfs_super_num_devices(disk_super),
-		       (unsigned long long)fs_devices->open_devices);
-		if (btrfs_test_opt(tree_root, DEGRADED))
-			printk("continuing in degraded mode\n");
-		else {
-			goto fail_sb_buffer;
-		}
-	}
-
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
 				    4 * 1024 * 1024 / PAGE_CACHE_SIZE);
@@ -1672,7 +1679,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	mutex_lock(&fs_info->chunk_mutex);
 	ret = btrfs_read_chunk_tree(chunk_root);
 	mutex_unlock(&fs_info->chunk_mutex);
-	BUG_ON(ret);
+	if (ret) {
+		printk("btrfs: failed to read chunk tree on %s\n", sb->s_id);
+		goto fail_chunk_root;
+	}
 
 	btrfs_close_extra_devices(fs_devices);
 
@@ -1684,7 +1694,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 					  btrfs_super_root(disk_super),
 					  blocksize, generation);
 	if (!tree_root->node)
-		goto fail_sb_buffer;
+		goto fail_chunk_root;
 
 
 	ret = find_and_setup_root(tree_root, fs_info,
@@ -1753,6 +1763,8 @@ fail_extent_root:
 	free_extent_buffer(extent_root->node);
 fail_tree_root:
 	free_extent_buffer(tree_root->node);
+fail_chunk_root:
+	free_extent_buffer(chunk_root->node);
 fail_sys_array:
 fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->fixup_workers);
@@ -1823,9 +1835,10 @@ int write_all_supers(struct btrfs_root *root)
 			total_errors++;
 			continue;
 		}
-		if (!dev->in_fs_metadata)
+		if (!dev->in_fs_metadata || !dev->writeable)
 			continue;
 
+		btrfs_set_stack_device_generation(dev_item, 0);
 		btrfs_set_stack_device_type(dev_item, dev->type);
 		btrfs_set_stack_device_id(dev_item, dev->devid);
 		btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
@@ -1834,6 +1847,7 @@ int write_all_supers(struct btrfs_root *root)
 		btrfs_set_stack_device_io_width(dev_item, dev->io_width);
 		btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
 		memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
+		memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
 		flags = btrfs_super_flags(sb);
 		btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
 
@@ -1881,7 +1895,7 @@ int write_all_supers(struct btrfs_root *root)
 		dev = list_entry(cur, struct btrfs_device, dev_list);
 		if (!dev->bdev)
 			continue;
-		if (!dev->in_fs_metadata)
+		if (!dev->in_fs_metadata || !dev->writeable)
 			continue;
 
 		BUG_ON(!dev->pending_io);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index af2de30dbea..197422c1dc4 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -355,7 +355,7 @@ __btrfs_find_block_group(struct btrfs_root *root,
 	if (search_start) {
 		struct btrfs_block_group_cache *shint;
 		shint = btrfs_lookup_first_block_group(info, search_start);
-		if (shint && block_group_bits(shint, data) && !shint->ro) {
+		if (shint && block_group_bits(shint, data)) {
 			spin_lock(&shint->lock);
 			used = btrfs_block_group_used(&shint->item);
 			if (used + shint->pinned + shint->reserved <
@@ -366,7 +366,7 @@ __btrfs_find_block_group(struct btrfs_root *root,
 			spin_unlock(&shint->lock);
 		}
 	}
-	if (hint && !hint->ro && block_group_bits(hint, data)) {
+	if (hint && block_group_bits(hint, data)) {
 		spin_lock(&hint->lock);
 		used = btrfs_block_group_used(&hint->item);
 		if (used + hint->pinned + hint->reserved <
@@ -392,7 +392,7 @@ again:
 		last = cache->key.objectid + cache->key.offset;
 		used = btrfs_block_group_used(&cache->item);
 
-		if (!cache->ro && block_group_bits(cache, data)) {
+		if (block_group_bits(cache, data)) {
 			free_check = div_factor(cache->key.offset, factor);
 			if (used + cache->pinned + cache->reserved <
 			    free_check) {
@@ -1843,9 +1843,9 @@ static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
 	spin_unlock(&cache->space_info->lock);
 }
 
-static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags)
+u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
 {
-	u64 num_devices = root->fs_info->fs_devices->num_devices;
+	u64 num_devices = root->fs_info->fs_devices->rw_devices;
 
 	if (num_devices == 1)
 		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
@@ -1877,13 +1877,11 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_space_info *space_info;
 	u64 thresh;
-	u64 start;
-	u64 num_bytes;
 	int ret = 0;
 
 	mutex_lock(&extent_root->fs_info->chunk_mutex);
 
-	flags = reduce_alloc_profile(extent_root, flags);
+	flags = btrfs_reduce_alloc_profile(extent_root, flags);
 
 	space_info = __find_space_info(extent_root->fs_info, flags);
 	if (!space_info) {
@@ -1913,16 +1911,11 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans,
 	}
 	spin_unlock(&space_info->lock);
 
-	ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
+	ret = btrfs_alloc_chunk(trans, extent_root, flags);
 	if (ret) {
 printk("space info full %Lu\n", flags);
 		space_info->full = 1;
-		goto out;
 	}
-
-	ret = btrfs_make_block_group(trans, extent_root, 0, flags,
-		     BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
-	BUG_ON(ret);
 out:
 	mutex_unlock(&extent_root->fs_info->chunk_mutex);
 	return ret;
@@ -3040,7 +3033,7 @@ static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
 		data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
 	}
 again:
-	data = reduce_alloc_profile(root, data);
+	data = btrfs_reduce_alloc_profile(root, data);
 	/*
 	 * the only place that sets empty_size is btrfs_realloc_node, which
 	 * is not called recursively on allocations
@@ -5136,7 +5129,8 @@ static int noinline relocate_one_path(struct btrfs_trans_handle *trans,
 			else
 				btrfs_node_key_to_cpu(eb, &keys[level], 0);
 		}
-		if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+		if (nodes[0] &&
+		    ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
 			eb = path->nodes[0];
 			ret = replace_extents_in_leaf(trans, reloc_root, eb,
 						      group, reloc_inode);
@@ -5377,7 +5371,7 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
 	u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
 		BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
 
-	num_devices = root->fs_info->fs_devices->num_devices;
+	num_devices = root->fs_info->fs_devices->rw_devices;
 	if (num_devices == 1) {
 		stripped |= BTRFS_BLOCK_GROUP_DUP;
 		stripped = flags & ~stripped;
@@ -5801,6 +5795,8 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		BUG_ON(ret);
 
 		set_avail_alloc_bits(root->fs_info, cache->flags);
+		if (btrfs_chunk_readonly(root, cache->key.objectid))
+			set_block_group_readonly(cache);
 	}
 	ret = 0;
 error:
@@ -5889,6 +5885,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	block_group->space_info->total_bytes -= block_group->key.offset;
 	block_group->space_info->bytes_readonly -= block_group->key.offset;
 	spin_unlock(&block_group->space_info->lock);
+	block_group->space_info->full = 0;
 
 	/*
 	memset(shrink_block_group, 0, sizeof(*shrink_block_group));
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 52863cebd59..f43df72b0e1 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -405,7 +405,7 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
 		devid = simple_strtoull(devstr, &end, 10);
 		printk(KERN_INFO "resizing devid %llu\n", devid);
 	}
-	device = btrfs_find_device(root, devid, NULL);
+	device = btrfs_find_device(root, devid, NULL, NULL);
 	if (!device) {
 		printk(KERN_INFO "resizer unable to find device %llu\n", devid);
 		ret = -EINVAL;
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 04a3bf81650..92393cc60d0 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -431,6 +431,11 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 	if (error)
 		goto error_free_subvol_name;
 
+	if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
+		error = -EACCES;
+		goto error_close_devices;
+	}
+
 	bdev = fs_devices->latest_bdev;
 	s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 	if (IS_ERR(s))
@@ -444,6 +449,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 			goto error_close_devices;
 		}
 
+		btrfs_close_devices(fs_devices);
 	} else {
 		char b[BDEVNAME_SIZE];
 
@@ -512,6 +518,9 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 		ret =  btrfs_commit_super(root);
 		WARN_ON(ret);
 	} else {
+		if (root->fs_info->fs_devices->rw_devices == 0)
+			return -EACCES;
+
 		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 			return -EINVAL;
 
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 80a27284dbf..d6f1996de62 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -40,6 +40,12 @@ struct map_lookup {
 	struct btrfs_bio_stripe stripes[];
 };
 
+static int init_first_rw_device(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				struct btrfs_device *device);
+static int btrfs_relocate_sys_chunks(struct btrfs_root *root);
+
+
 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
 			    (sizeof(struct btrfs_bio_stripe) * (n)))
 
@@ -69,25 +75,31 @@ static void unlock_chunks(struct btrfs_root *root)
 int btrfs_cleanup_fs_uuids(void)
 {
 	struct btrfs_fs_devices *fs_devices;
-	struct list_head *uuid_cur;
-	struct list_head *devices_cur;
 	struct btrfs_device *dev;
 
-	list_for_each(uuid_cur, &fs_uuids) {
-		fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices,
-					list);
+	while (!list_empty(&fs_uuids)) {
+		fs_devices = list_entry(fs_uuids.next,
+					struct btrfs_fs_devices, list);
+		list_del(&fs_devices->list);
 		while(!list_empty(&fs_devices->devices)) {
-			devices_cur = fs_devices->devices.next;
-			dev = list_entry(devices_cur, struct btrfs_device,
-					 dev_list);
+			dev = list_entry(fs_devices->devices.next,
+					 struct btrfs_device, dev_list);
 			if (dev->bdev) {
 				close_bdev_excl(dev->bdev);
 				fs_devices->open_devices--;
 			}
+			fs_devices->num_devices--;
+			if (dev->writeable)
+				fs_devices->rw_devices--;
 			list_del(&dev->dev_list);
+			list_del(&dev->dev_alloc_list);
 			kfree(dev->name);
 			kfree(dev);
 		}
+		WARN_ON(fs_devices->num_devices);
+		WARN_ON(fs_devices->open_devices);
+		WARN_ON(fs_devices->rw_devices);
+		kfree(fs_devices);
 	}
 	return 0;
 }
@@ -257,6 +269,9 @@ static noinline int device_list_add(const char *path,
 				       disk_super->dev_item.uuid);
 	}
 	if (!device) {
+		if (fs_devices->opened)
+			return -EBUSY;
+
 		device = kzalloc(sizeof(*device), GFP_NOFS);
 		if (!device) {
 			/* we can safely leave the fs_devices entry around */
@@ -273,8 +288,9 @@ static noinline int device_list_add(const char *path,
 			kfree(device);
 			return -ENOMEM;
 		}
+		INIT_LIST_HEAD(&device->dev_alloc_list);
 		list_add(&device->dev_list, &fs_devices->devices);
-		list_add(&device->dev_alloc_list, &fs_devices->alloc_list);
+		device->fs_devices = fs_devices;
 		fs_devices->num_devices++;
 	}
 
@@ -288,58 +304,94 @@ static noinline int device_list_add(const char *path,
 
 int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
 {
-	struct list_head *head = &fs_devices->devices;
+	struct list_head *tmp;
 	struct list_head *cur;
 	struct btrfs_device *device;
+	int seed_devices = 0;
 
 	mutex_lock(&uuid_mutex);
 again:
-	list_for_each(cur, head) {
+	list_for_each_safe(cur, tmp, &fs_devices->devices) {
 		device = list_entry(cur, struct btrfs_device, dev_list);
-		if (!device->in_fs_metadata) {
-			struct block_device *bdev;
-			list_del(&device->dev_list);
-			list_del(&device->dev_alloc_list);
+		if (device->in_fs_metadata)
+			continue;
+
+		if (device->bdev) {
+			close_bdev_excl(device->bdev);
+			device->bdev = NULL;
+			fs_devices->open_devices--;
+		}
+		if (device->writeable) {
+			list_del_init(&device->dev_alloc_list);
+			device->writeable = 0;
+			fs_devices->rw_devices--;
+		}
+		if (!seed_devices) {
+			list_del_init(&device->dev_list);
 			fs_devices->num_devices--;
-			if (device->bdev) {
-				bdev = device->bdev;
-				fs_devices->open_devices--;
-				mutex_unlock(&uuid_mutex);
-				close_bdev_excl(bdev);
-				mutex_lock(&uuid_mutex);
-			}
 			kfree(device->name);
 			kfree(device);
-			goto again;
 		}
 	}
+
+	if (fs_devices->seed) {
+		fs_devices = fs_devices->seed;
+		seed_devices = 1;
+		goto again;
+	}
+
 	mutex_unlock(&uuid_mutex);
 	return 0;
 }
 
-int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
+static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 {
-	struct list_head *head = &fs_devices->devices;
+	struct btrfs_fs_devices *seed_devices;
 	struct list_head *cur;
 	struct btrfs_device *device;
+again:
+	if (--fs_devices->opened > 0)
+		return 0;
 
-	mutex_lock(&uuid_mutex);
-	list_for_each(cur, head) {
+	list_for_each(cur, &fs_devices->devices) {
 		device = list_entry(cur, struct btrfs_device, dev_list);
 		if (device->bdev) {
 			close_bdev_excl(device->bdev);
 			fs_devices->open_devices--;
 		}
+		if (device->writeable) {
+			list_del_init(&device->dev_alloc_list);
+			fs_devices->rw_devices--;
+		}
+
 		device->bdev = NULL;
+		device->writeable = 0;
 		device->in_fs_metadata = 0;
 	}
-	fs_devices->mounted = 0;
-	mutex_unlock(&uuid_mutex);
+	fs_devices->opened = 0;
+	fs_devices->seeding = 0;
+	fs_devices->sprouted = 0;
+
+	seed_devices = fs_devices->seed;
+	fs_devices->seed = NULL;
+	if (seed_devices) {
+		fs_devices = seed_devices;
+		goto again;
+	}
 	return 0;
 }
 
-int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
-		       int flags, void *holder)
+int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
+{
+	int ret;
+
+	mutex_lock(&uuid_mutex);
+	ret = __btrfs_close_devices(fs_devices);
+	mutex_unlock(&uuid_mutex);
+	return ret;
+}
+
+int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices, void *holder)
 {
 	struct block_device *bdev;
 	struct list_head *head = &fs_devices->devices;
@@ -350,24 +402,18 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 	struct btrfs_super_block *disk_super;
 	u64 latest_devid = 0;
 	u64 latest_transid = 0;
-	u64 transid;
 	u64 devid;
+	int seeding = 1;
 	int ret = 0;
 
-	mutex_lock(&uuid_mutex);
-	if (fs_devices->mounted)
-		goto out;
-
 	list_for_each(cur, head) {
 		device = list_entry(cur, struct btrfs_device, dev_list);
 		if (device->bdev)
 			continue;
-
 		if (!device->name)
 			continue;
 
-		bdev = open_bdev_excl(device->name, flags, holder);
-
+		bdev = open_bdev_excl(device->name, MS_RDONLY, holder);
 		if (IS_ERR(bdev)) {
 			printk("open %s failed\n", device->name);
 			goto error;
@@ -387,16 +433,32 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		if (devid != device->devid)
 			goto error_brelse;
 
-		transid = btrfs_super_generation(disk_super);
-		if (!latest_transid || transid > latest_transid) {
+		if (memcmp(device->uuid, disk_super->dev_item.uuid,
+			   BTRFS_UUID_SIZE))
+			goto error_brelse;
+
+		device->generation = btrfs_super_generation(disk_super);
+		if (!latest_transid || device->generation > latest_transid) {
 			latest_devid = devid;
-			latest_transid = transid;
+			latest_transid = device->generation;
 			latest_bdev = bdev;
 		}
 
+		if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_SEEDING) {
+			device->writeable = 0;
+		} else {
+			device->writeable = !bdev_read_only(bdev);
+			seeding = 0;
+		}
+
 		device->bdev = bdev;
 		device->in_fs_metadata = 0;
 		fs_devices->open_devices++;
+		if (device->writeable) {
+			fs_devices->rw_devices++;
+			list_add(&device->dev_alloc_list,
+				 &fs_devices->alloc_list);
+		}
 		continue;
 
 error_brelse:
@@ -410,11 +472,32 @@ error:
 		ret = -EIO;
 		goto out;
 	}
-	fs_devices->mounted = 1;
+	fs_devices->seeding = seeding;
+	fs_devices->opened = 1;
 	fs_devices->latest_bdev = latest_bdev;
 	fs_devices->latest_devid = latest_devid;
 	fs_devices->latest_trans = latest_transid;
+	fs_devices->total_rw_bytes = 0;
 out:
+	return ret;
+}
+
+int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
+		       int flags, void *holder)
+{
+	int ret;
+
+	mutex_lock(&uuid_mutex);
+	if (fs_devices->opened) {
+		if (fs_devices->sprouted) {
+			ret = -EBUSY;
+		} else {
+			fs_devices->opened++;
+			ret = 0;
+		}
+	} else {
+		ret = __btrfs_open_devices(fs_devices, holder);
+	}
 	mutex_unlock(&uuid_mutex);
 	return ret;
 }
@@ -481,12 +564,12 @@ error:
  */
 static noinline int find_free_dev_extent(struct btrfs_trans_handle *trans,
 					 struct btrfs_device *device,
-					 struct btrfs_path *path,
 					 u64 num_bytes, u64 *start)
 {
 	struct btrfs_key key;
 	struct btrfs_root *root = device->dev_root;
 	struct btrfs_dev_extent *dev_extent = NULL;
+	struct btrfs_path *path;
 	u64 hole_size = 0;
 	u64 last_byte = 0;
 	u64 search_start = 0;
@@ -496,8 +579,11 @@ static noinline int find_free_dev_extent(struct btrfs_trans_handle *trans,
 	int start_found;
 	struct extent_buffer *l;
 
-	start_found = 0;
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
 	path->reada = 2;
+	start_found = 0;
 
 	/* FIXME use last free of some kind */
 
@@ -581,7 +667,6 @@ check_pending:
 	/* we have to make sure we didn't find an extent that has already
 	 * been allocated by the map tree or the original allocation
 	 */
-	btrfs_release_path(root, path);
 	BUG_ON(*start < search_start);
 
 	if (*start + num_bytes > search_end) {
@@ -589,10 +674,10 @@ check_pending:
 		goto error;
 	}
 	/* check for pending inserts here */
-	return 0;
+	ret = 0;
 
 error:
-	btrfs_release_path(root, path);
+	btrfs_free_path(path);
 	return ret;
 }
 
@@ -644,11 +729,10 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-int noinline btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
+int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 			   struct btrfs_device *device,
 			   u64 chunk_tree, u64 chunk_objectid,
-			   u64 chunk_offset,
-			   u64 num_bytes, u64 *start)
+			   u64 chunk_offset, u64 start, u64 num_bytes)
 {
 	int ret;
 	struct btrfs_path *path;
@@ -662,13 +746,8 @@ int noinline btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	ret = find_free_dev_extent(trans, device, path, num_bytes, start);
-	if (ret) {
-		goto err;
-	}
-
 	key.objectid = device->devid;
-	key.offset = *start;
+	key.offset = start;
 	key.type = BTRFS_DEV_EXTENT_KEY;
 	ret = btrfs_insert_empty_item(trans, root, path, &key,
 				      sizeof(*extent));
@@ -687,7 +766,6 @@ int noinline btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 
 	btrfs_set_dev_extent_length(leaf, extent, num_bytes);
 	btrfs_mark_buffer_dirty(leaf);
-err:
 	btrfs_free_path(path);
 	return ret;
 }
@@ -735,12 +813,18 @@ error:
 	return ret;
 }
 
-static noinline int find_next_devid(struct btrfs_root *root,
-				    struct btrfs_path *path, u64 *objectid)
+static noinline int find_next_devid(struct btrfs_root *root, u64 *objectid)
 {
 	int ret;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
+	struct btrfs_path *path;
+
+	root = root->fs_info->chunk_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
 
 	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
 	key.type = BTRFS_DEV_ITEM_KEY;
@@ -763,7 +847,7 @@ static noinline int find_next_devid(struct btrfs_root *root,
 	}
 	ret = 0;
 error:
-	btrfs_release_path(root, path);
+	btrfs_free_path(path);
 	return ret;
 }
 
@@ -781,7 +865,6 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 	struct btrfs_key key;
 	unsigned long ptr;
-	u64 free_devid = 0;
 
 	root = root->fs_info->chunk_root;
 
@@ -789,13 +872,9 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	ret = find_next_devid(root, path, &free_devid);
-	if (ret)
-		goto out;
-
 	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
 	key.type = BTRFS_DEV_ITEM_KEY;
-	key.offset = free_devid;
+	key.offset = device->devid;
 
 	ret = btrfs_insert_empty_item(trans, root, path, &key,
 				      sizeof(*dev_item));
@@ -805,8 +884,8 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 	leaf = path->nodes[0];
 	dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
 
-	device->devid = free_devid;
 	btrfs_set_device_id(leaf, dev_item, device->devid);
+	btrfs_set_device_generation(leaf, dev_item, 0);
 	btrfs_set_device_type(leaf, dev_item, device->type);
 	btrfs_set_device_io_align(leaf, dev_item, device->io_align);
 	btrfs_set_device_io_width(leaf, dev_item, device->io_width);
@@ -819,9 +898,11 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 
 	ptr = (unsigned long)btrfs_device_uuid(dev_item);
 	write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
+	ptr = (unsigned long)btrfs_device_fsid(dev_item);
+	write_extent_buffer(leaf, root->fs_info->fsid, ptr, BTRFS_UUID_SIZE);
 	btrfs_mark_buffer_dirty(leaf);
-	ret = 0;
 
+	ret = 0;
 out:
 	btrfs_free_path(path);
 	return ret;
@@ -832,11 +913,7 @@ static int btrfs_rm_dev_item(struct btrfs_root *root,
 {
 	int ret;
 	struct btrfs_path *path;
-	struct block_device *bdev = device->bdev;
-	struct btrfs_device *next_dev;
 	struct btrfs_key key;
-	u64 total_bytes;
-	struct btrfs_fs_devices *fs_devices;
 	struct btrfs_trans_handle *trans;
 
 	root = root->fs_info->chunk_root;
@@ -863,25 +940,6 @@ static int btrfs_rm_dev_item(struct btrfs_root *root,
 	ret = btrfs_del_item(trans, root, path);
 	if (ret)
 		goto out;
-
-	/*
-	 * at this point, the device is zero sized.  We want to
-	 * remove it from the devices list and zero out the old super
-	 */
-	list_del_init(&device->dev_list);
-	list_del_init(&device->dev_alloc_list);
-	fs_devices = root->fs_info->fs_devices;
-
-	next_dev = list_entry(fs_devices->devices.next, struct btrfs_device,
-			      dev_list);
-	if (bdev == root->fs_info->sb->s_bdev)
-		root->fs_info->sb->s_bdev = next_dev->bdev;
-	if (bdev == fs_devices->latest_bdev)
-		fs_devices->latest_bdev = next_dev->bdev;
-
-	total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
-	btrfs_set_super_num_devices(&root->fs_info->super_copy,
-				    total_bytes - 1);
 out:
 	btrfs_free_path(path);
 	unlock_chunks(root);
@@ -892,11 +950,14 @@ out:
 int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 {
 	struct btrfs_device *device;
+	struct btrfs_device *next_device;
 	struct block_device *bdev;
 	struct buffer_head *bh = NULL;
 	struct btrfs_super_block *disk_super;
 	u64 all_avail;
 	u64 devid;
+	u64 num_devices;
+	u8 *dev_uuid;
 	int ret = 0;
 
 	mutex_lock(&uuid_mutex);
@@ -907,14 +968,14 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		root->fs_info->avail_metadata_alloc_bits;
 
 	if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) &&
-	    btrfs_super_num_devices(&root->fs_info->super_copy) <= 4) {
+	    root->fs_info->fs_devices->rw_devices <= 4) {
 		printk("btrfs: unable to go below four devices on raid10\n");
 		ret = -EINVAL;
 		goto out;
 	}
 
 	if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) &&
-	    btrfs_super_num_devices(&root->fs_info->super_copy) <= 2) {
+	    root->fs_info->fs_devices->rw_devices <= 2) {
 		printk("btrfs: unable to go below two devices on raid1\n");
 		ret = -EINVAL;
 		goto out;
@@ -941,15 +1002,15 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 			printk("btrfs: no missing devices found to remove\n");
 			goto out;
 		}
-
 	} else {
-		bdev = open_bdev_excl(device_path, 0,
+		bdev = open_bdev_excl(device_path, MS_RDONLY,
 				      root->fs_info->bdev_holder);
 		if (IS_ERR(bdev)) {
 			ret = PTR_ERR(bdev);
 			goto out;
 		}
 
+		set_blocksize(bdev, 4096);
 		bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
 		if (!bh) {
 			ret = -EIO;
@@ -957,45 +1018,97 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		}
 		disk_super = (struct btrfs_super_block *)bh->b_data;
 		if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-		    sizeof(disk_super->magic))) {
-			ret = -ENOENT;
-			goto error_brelse;
-		}
-		if (memcmp(disk_super->fsid, root->fs_info->fsid,
-			   BTRFS_FSID_SIZE)) {
+			    sizeof(disk_super->magic))) {
 			ret = -ENOENT;
 			goto error_brelse;
 		}
 		devid = le64_to_cpu(disk_super->dev_item.devid);
-		device = btrfs_find_device(root, devid, NULL);
+		dev_uuid = disk_super->dev_item.uuid;
+		device = btrfs_find_device(root, devid, dev_uuid,
+					   disk_super->fsid);
 		if (!device) {
 			ret = -ENOENT;
 			goto error_brelse;
 		}
+	}
 
+	if (device->writeable && root->fs_info->fs_devices->rw_devices == 1) {
+		printk("btrfs: unable to remove the only writeable device\n");
+		ret = -EINVAL;
+		goto error_brelse;
+	}
+
+	if (device->writeable) {
+		list_del_init(&device->dev_alloc_list);
+		root->fs_info->fs_devices->rw_devices--;
 	}
-	root->fs_info->fs_devices->num_devices--;
-	root->fs_info->fs_devices->open_devices--;
 
 	ret = btrfs_shrink_device(device, 0);
 	if (ret)
 		goto error_brelse;
 
-
 	ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device);
 	if (ret)
 		goto error_brelse;
 
-	if (bh) {
+	device->in_fs_metadata = 0;
+	if (device->fs_devices == root->fs_info->fs_devices) {
+		list_del_init(&device->dev_list);
+		root->fs_info->fs_devices->num_devices--;
+		if (device->bdev)
+			device->fs_devices->open_devices--;
+	}
+
+	next_device = list_entry(root->fs_info->fs_devices->devices.next,
+				 struct btrfs_device, dev_list);
+	if (device->bdev == root->fs_info->sb->s_bdev)
+		root->fs_info->sb->s_bdev = next_device->bdev;
+	if (device->bdev == root->fs_info->fs_devices-&